Concurrent error detection (CED) is the detection of errors or faults in a circuit or data path concurrent with normal operation of that circuit. The general approach for CED is to calculate a check symbol for the inputs to the circuit under operation, predict the check symbol that will result for the output of the circuit for those inputs, and compare the predicted check symbol to the one that is actually calculated for the output. If the predicted and actual check symbols are different, an error or fault has been detected. The alternative to this check symbol prediction is to use a second copy of the circuit under operation and compare the results of the two circuits. For some classes of circuits the prediction of the output check symbol can require less circuitry than a second copy of the circuit being tested. Four examples of these types of circuits are examined in this dissertation: Arithmetic Logic Units (ALUs), array multipliers, self-synchronous scrambler-descrambler pairs with their intervening data path, and switch fabrics. Faults in integrated circuits tend to produce unidirectional errors. Unidirectional errors are those in which all of the errors are in the same direction (e.g., 0 to 1 errors) within the block of data covered by a given check symbol. For this reason, codes that are optimized for unidirectional errors are the focus of investigation for most of the applications. In particular, the Bose-Lin codes are examined for those applications where unidirectional errors are expected to be typical. In order to examine the performance of the Bose-Lin codes in one of these applications, it was necessary to determine the theoretical performance for Bose- Lin codes for error rates beyond what had been previously studied. This analysis of Bose-Lin codes with large numbers of "burst" errors also included a further generalization of the codes. / Graduation date: 2002